Neuro - Genetic approaches to classification of Face Images with effective feature selection using hybrid classifiers

Abstract

The ever-increasing volume in the collection of image data in various fields of science, medicine, security and other fields has brought the necessity to extract knowledge. Face classification/recognition is one of the challenging problems of computer vision. The use of Data mining techniques has a legitimate and enabling ways to explore these large image collections using neuro- genetic approaches. A novel Symmetric Based Algorithm is proposed for face detection in still gray level images, which acts as a selective attentional mechanism. The three face classifiers/recognizers, Linear Discriminant Analysis (LDA), Line Based Algorithm (LBA) and Kernel Direct Discriminant Analysis (KDDA) are fused using Radial Basis network for efficient feature extraction of the face images. The use of Genetic algorithm approach optimizes the weights of neural network to extract only the essential features that effectively and successively improves the classification/recognition accuracy. A total of 1024 images for 22 subjects taken from BioID Laboratory, Texas, USA are used for analysis.